1. Field of the Invention
The present invention relates to a Metal-Insulator-Semiconductor (hereinafter referred to as MIS) type semiconductor device, and specifically relates to a MIS transistor. More specifically the present invention relates to a thin-film state MIS type semiconductor device, a thin-film transistor (hereinafter referred to as TFT) formed on an insulating substrate, and above all relates to a MIS type semiconductor device having a so-called inverse stagger type structure, a channel forming region of which is positioned above a gate electrode. The application fields of the present invention are semiconductor integrated circuit formed on an insulating substrate, for example an active matrix type circuit used in a liquid crystal display, a driving circuit for an image sensor and the like.
2. Description of the Related Art
In recent years, a thin-film state MIS type semiconductor device formed on an insulating substrate has been used. For example an active matrix type liquid crystal display etc. have been known. There are two kinds of active matrix type circuit which are on the present market, one is a TFT using type and the other is such diode as a MIM (Metal Insulator Metal) using type. Particularly, the former has been successfully manufactured recently because it has given a high quality image.
As an active matrix circuit using TFT, two sorts of TFT, which employ such polycrystal semiconductor as polycrystal silicon and such amorphous semiconductor as amorphous silicon, have been known. The former has a difficulty in preparation for a large picture owing to its preparing process, and then the latter, which can be prepared at a temperature of 350xc2x0 C. or below, has been mainly used for a large picture.
FIG. 2 shows a preparing process of a conventional amorphous silicon TFT (an inverse stagger type). As the substrate 201, such heat resistant and non-alkali glass as Corning 7059 is used. Since the maximum temperature of preparing process for the amorphous silicon TFT is around 350xc2x0 C., it is required that the material therefor be resistant up to this temperature. Especially in case of a liquid crystal display panel, the material is in need of a heat resistant property and a high glass transition temperature not to be distorted by heat treatment. The Corning 7059 can meet for the need, as the glass transition temperature thereof is a little lower than 600xc2x0 C.
Also, in order to stabilize an operation of TFT, it is not desirable that such movable ion as Na is contained in a substrate. There is no problem with the Corning 7059 having quite low content of alkalies, but it is necessary to form such passivation film as a silicon nitride film or an aluminum oxide film so that the movable ions in a substrate do not invade into TFT, in case a lot of Na etc. is contained in the substrate.
Next, a film is formed using materials such as aluminum or tantalum, a patterning is effected using the mask 1 and the gate electrode 202 is formed. In particular, in order to prevent a short circuit between the gate electrode wiring and the upper wiring thereof, it is recommended to form the oxide film 203 on this gate electrode surface. An anode oxidation method is chiefly employed in the forming way of the oxide film.
Then, the gate insulating film 204 is formed. As the gate insulating film, a silicon nitride is generally used. But a silicon oxide can be also used, and a silicide mixed with nitrogen and oxygen in a free amount ratio can be used. Further both of the single layer and the polylayer films can be available. When a silicon nitride film is used as the gate insulating film and a plasma CVD method is applied, the process temperature becomes 350xc2x0 C., the maximum temperature of the process, the situation of which is shown in FIG. 2(A).
Next, an amorphous silicon film is formed. it is necessary to raise a temperature of the substrate up to 250 to 300xc2x0 C., in case a plasma CVD method is used. The film thickness is desired as thin as possible, it is generally to be 10 to 100 nm, and preferably to be 10 to 30 nm. The amorphous silicon region 205 is formed by patterning with the mask 2 which later will become a channel forming region of TFT. The situation so far is shown in FIG. 2(B).
Further, a silicon nitride film is formed all over the surface, which is patterned with the mask 3 to get the etching stopper 206. This stopper is prepared so that the amorphous silicon region 205 for a channel forming region will not be etched by mistake in a later process, because the amorphous silicon region 205 is as thin as 10 to 100 nm as aforementioned. Also, since the amorphous silicon region under part of the etching stopper functions as a channel forming region, the etching stopper is designed to overlap with the gate electrode as fully as possible. A conventional mask alignment, however, gives a somewhat discrepancy, then the patterning is carried out so as to achieve an enough overlapping with the gate electrode.
After that, N-type or P-type conductivity silicon film is formed. An ordinary amorphous silicon TFT is N-type one. This silicon film is prepared to be a microcrystal state, as an amorphous silicon is too low in a conductivity. N-type microcrystal silicon film can be prepared at a temperature of 350xc2x0 C. or below by a plasma CVD method. But it is not low enough in resistance, and then the thickness of 200 nm or more was required. P-type microcrystal silicon film was too big in resistance to be used, it was, therefore, difficult to prepare a P-channel type TFT with an amorphous silicon.
The silicon film prepared in this way is patterned using the mask 4 to form N-type microcrystal silicon region 207. The situation so far is shown in FIG. 2(C), in which a function of TFT can not be realized, as (N-type) microcrystal silicon film is connected with itself on the etching stopper. It is, therefore, necessary to separate the connection, and the groove 208 is formed by separating the connection using the mask 5. It is feared that even the amorphous silicon region 205 as a base is etched out making a mistake, if there is no etching stopper in this case. This is caused by the reason that the microcrystal silicon region 207 is from several to some ten times or more as thick as the amorphous silicon region thereunder. Afterward, the wiring 209 and the pixel electrode 210 are prepared, using the masks 6 and 7 by a known method, the situation of which is shown in FIG. 2(D).
In the method stated above, so many seven sheets of mask are used, and it is worried that a yield will be lowered. Therefore, a decreasing way of the number of masks has been proposed as will be described in the following; Firstly, a gate electrode part is patterned on a substrate using the first mask. Then, a gate insulating film is formed, further, an amorphous silicon film and a silicon nitride film (later will be an etching stopper) are continuously formed. Next, an etching stopper is formed by etching only the silicon nitride film in a way of self-alignment, using a gate electrode as a mask and exposing from the back. Then, thereon a microcrystal silicon film is formed, and a TFT region including a groove over the channel (corresponds to the numeral 208 in FIG. 2) is formed using the second mask. Afterward, a wiring and an electrode are formed using the third and fourth masks. Finally, the equivalent one as shown in FIG. 2(D) can be obtained. In such way, the number of masks can be reduced by three utilizing a self-alignment process.
Thus formed TFT is very uneven, as distinct from FIG. 2(D). This is mainly derived from the gate electrode part (contains gate electrode oxide 203), the etching stopper, and the microcrystal silicon region. A total 800 nm of unevenness will happen, supposing that each thickness of the gate electrode part, the etching stopper, and the microcrystal silicon region 206 is 300 nm, 200 nm, and 300 nm respectively. For example in case where TFT is used in an active matrix circuit of a liquid crystal display panel, the thickness of a cell is 5 to 6 xcexcm and is controlled to an accuracy of 0.1 xcexcm or less. Under such condition, even 1 xcexcm of unevenness will cause a remarkable defect to a thickness uniformity of a cell.
However, any of these factors to cause the unevenness of TFT can not be cut down easily. Namely, the thinner a gate electrode part is, the higher the resistance of gate electrode wiring will be. On the other hand, in order to keep the constant resistance, widening of a gate electrode (i.e. lengthening of a channel) will bring about not only lowering of TFT operation speed but also large area of TFT part, which will cause an aperture ratio to be lowered in case TFT is used in a liquid crystal display.
Also, in case the etching stopper is thin, there is a possibility that even an amorphous silicon region underlying a microcrystal silicon region will be in error to be etched, during the etching of the microcrystal silicon region, and then a yield will be lowered. Further, in case the microcrystal silicon region is thin, the source/drain region resistance of TFT will be increased and ON/OFF ratio of TFT will be decreased.
Still more, the etching stopper will remain as it is, at the time of completion of TFT. The silicon nitride film used for the etching stopper has a nature to trap electric charge. If electric charge is trapped therein for some reason, an unwilling channel will be formed in the amorphous silicon region 205 thereunder, which will cause a leak of drain current. To avoid this problem, it is necessary to cause the etching stopper to be two layer structures of silicon oxide and silicon nitride. In this case, it is needed that the silicon oxide film be also thick enough, and preferably be 100 nm thick or more.
The present invention seeks to solve such problems. Accordingly, it is an object of the present invention to simplify the process. For example, it is to improve a yield by decreasing the number of masks compared with that in a conventional method. It is also to improve a through-put and to reduce a cost, by decreasing the film-forming processes. It is another object of the present invention to make TFT flat. According to this flattening, the problem posed in case TFT is used in a liquid crystal display panel can be settled. Also the flattening is an important technical subject in other applications, and can be applicable to what a conventional TFT has not been put to a practical use.
To overcome the above-mentioned problems, the present invention provides a new preparing method for TFT using no etching stopper, and also provides TFT prepared by the method thereof. Further, for purpose of making a microcrystal silicon region (source/drain) thin, the resistance thereof will be sufficiently rendered low. Still more, in accordance with the present invention, a sheet of silicon film will be formed by no way of conventional two-stage processes, which are the forming of an amorphous silicon region (film) to become a channel forming region, and the forming of a microcrystal silicon region (film) to become a source/drain region. And this one sheet of silicon film will be reformed partly for a source/drain region and partly for a channel forming region.
In case of the through-put improvement, it is the most important subject to reduce the film formings. The film forming process takes much time and also requires an equal time for cleaning of a chamber innerside, thereby, the actual condition being that the film forming is carried out between the chamber cleanings, in the present semiconductor process where the purest environment is demanded. It is, therefore, necessary for the purpose of through-put improvement to form a thin-film than a thick-one, and to form a single layer film than a multilayer one. This means that the reduction of the film forming processes is most desirable.
TFT in one mode of the present invention is claimed as follows; Firstly it is an insulated gate field effect transistor (TFT) of an inverted stagger type. A gate insulating film is formed covering a gate electrode provided on an insulating substrate, and a semiconductor film is formed on the gate insulating film to a thickness of 10 to 100 nm. But the upper part thereof provided on the gate electrode is substantially intrinsic so as to function as a channel forming region. Other parts are N-type or P-type, and fulfill function as source/drain. Also the part functioning as a channel forming region can be taken to be amorphous, semi-amorphous, microcrystal, polycrystal, or an intermediate state thereof. In case where an off-current is suppressed, an amorphous is preferable. On the other hand, the region functioning as source/drain is polycrystal, semiamorphous, or microcrystal, each of which is low enough in resistance. Moreover, the present invention is characterized in that this region is formed by applying a laser annealing.
In the claimed invention, a mass production can be attained, as the film forming is all right by forming only one layer of a semiconductor film. Also, the unevenness of TFT can be reduced, if a conventional microcrystal silicon is not formed. Of course, it is needless to say that the present invention do not seek to form a channel forming region and such impurity region as source/drain with only one layer of a semiconductor film, but a multilayer of a semiconductor film may be formed, to improve the device properties taking cost and characteristics into account.
TFT in another mode of the present invention is characterized in that an etching stopper is not provided on a channel forming region, and in that at least silicon nitride or such similar material as have a nature to trap an electric charge does not exist, adhering to a channel forming region or through a thin (100 nm or less) insulating film.
The existence of an etching stopper is an important factor for the unevenness of TFT, and in case the stopper is composed of such material as silicon nitride, there will happen a leak of drain current. Such problems can be overcome by the above stated technological thoughts of the present invention. Of course, the technological thoughts of the present invention do not require that nobody exists on a channel forming region, and there poses no problem, even if somebody exists to such degree as not to reveal the above-mentioned problems.
The preparation of TFT according to the present invention will be effected by the method as shown in FIG. 1. Of course, there may be some modifications to the process in FIG. 1, if necessary. As shown in FIG. 1, the gate electrode 102 will be patterned with the mask 1, on the substrate 101 which is heat-resistant and non-alkali glass. If necessary, it may be possible that an insulating property is improved, by forming the oxide film 103 on the surface of gate electrode. Then, the gate insulating film 104 will be formed, thereby FIG. 1(A) being obtained.
Next, an amorphous, semiamorphous, microcrystal, polycrystal, or the intermediate state thin silicon film will be formed, and patterned with the mask 2 to form the semiconductor region 105. In many actual cases, an amorphous silicon film will be formed considering a film forming temperature and an off-current, but it may be possible that a polycrystal or a semiamorphous silicon is obtained, using such low temperature crystallization art as a laser annealing. In case a polycrystal or a semiamorphous silicon is used, a large electric field mobility can be obtained, however, it is not suitable for an active matrix circuit of liquid crystal display panel, because an off-current increases.
After that, a film to be a masking material against the laser beams, e.g., silicon nitride film rich in silicon content (50 nm or more of thickness is favorable) will be formed, which will be patterned with the mask 3. At this time, a photoresist can be made to remain on the silicon nitride film if necessary. Namely, in FIG. 1(C), the numeral 106 indicates a silicon nitride film and 107 indicates a photoresist. The thickness of photoresist is 100 nm or more and is favorably 500 nm or more, supposing a later ion implantation process.
At this situation, firstly an impurity will be selectively implanted into the semiconductor region 105, by an ion implantation or an ion doping method etc. thus being formed the impurity region 108. The impurity implantation, however, brings about very big defects in a semiconductor film, which no longer functions as a semiconductor. Accordingly, a crystallization will be effected by radiating laser beams from upward. In this laser annealing process, various conditions of silicon ranging from polycrystal but very close to single crystal to amorphous can be formed, by properly controlling a pulse width and an energy density of the laser beams.
If there is no silicon nitride film 106, the laser beams will reach the region where functions as a channel forming region not doped with an impurity, and will enable its part to be crystallized. In case where a silicon nitride film exists, by which a lot of beams will be absorbed, then the crystallization will not occur and the initial situation will be maintained. It is considered favorable from the mobility increasing point of view that a channel region will be crystallized by laser beams. The present-day laser technologies, however, have a deviation in a laser shot-energy, which gives rise to a remarkable deviation in the crystallization degree. As a result, TFT with different mobilities will be formed.
There is no problem when TFT with a constant mobility is only required. But the condition will be extremely strict, provided that the mobility is satisfied with a constant lower limit and further the off-current is satisfied with a constant upper limit. This comes from the reason that in general TFT with a big mobility will be also big in the off-current. For example, since not only the mobility but also the off-current is important factor, in an active matrix circuit of a liquid crystal display panel, uniformly good TFTs are required. Accordingly in such case, it is desired that TFT is formed using an amorphous silicon which has a low off-current, even if it has a low mobility, or using a material which is similar to the amorphous silicon. Consequently, in the present invention, the laser beams should be prohibited from being entered into a channel forming region by mistake, in the case of such purposes.
This doping process may be done by a laser doping. The laser doping is a method that a sample (a semiconductor film) will be placed in an atmosphere containing an impurity, and by irradiating laser beams or equivalent strong beams to the sample (semiconductor film) in the atmosphere, the sample surface will be heated and activated and an impurity gas will be decomposed to diffuse the impurity over the sample surface. By the laser doping, the impurity is introduced into the semiconductor film and the semiconductor film is crystallized. As the impurity gas, in general PH3 (phosphine) will be used in case where N-type is furnished, and B2H6 (diborane) will be used in case where P-type is furnished.
By the present inventors"" knowledge, in case of the laser beams radiation, a heating of a sample (the semiconductor film) at around 250 to 500xc2x0 C. enabled the impurity to diffuse into an inside of the sample, causing the impurity concentration thereof to be high enough. It is not desired to place the sample at a too high temperature, so as to keep an amorphous state in the channel forming region, and there is some restrictions added to a glass substrate. So that, it is preferable to put an end of heating to around 250 to 350xc2x0 C.
Also in case of the laser doping, the photoresist mask 107 is not always needed. In such doping method as an ion implantation, such mask as a photoresist, which is thick enough to enable an ion energy to be thoroughly slowed down, so that a high energy ion to be implanted will not enter into the channel forming region in mistake, is required. However, since the laser doping is a kind of heat-diffusion method, only such silicon nitride mask 106 as a material to have a full masking action to the heat-diffusion will do as a mask. The details about the laser doping technology are disclosed in the Japanese Patent Application No. Heisei 3-283981 filed by the present inventors et al.
After such doping was executed as above, the silicon nitride film 106 and the photoresist (will be evaporated by laser beams radiation in most cases) 107 will be removed, and then, the wiring 110 and the pixel electrode of ITO (Indium Tin Oxide) 111 will be formed using the mask 4 and the mask 5 by a known method. In the above processes, there totally used five sheets of mask, but these can be reduced to four sheets of mask, by freely using a self-alignment method as usual. Namely, for the formings of gate electrode, semiconductor region, as well as pixel electrode and wiring, 1-, 1-, and 2-sheet of mask are respectively necessary. The patterning for the silicon nitride film 106 etc. will be effected by the back exposure (e.g. by irradiating a light from under the substrate to the mask material (the silicon nitride film) provided on the semiconductor region), using the gate electrode as a mask.
As is evident in FIG. 1(D), TFT according to the present invention is smaller in an unevenness than that of a conventional one. This is due to the fact that the unevenness is chiefly caused by only an unevenness of the gate electrode. The semiconductor region 105 is very thin, and the thickness thereof is 10 to 100 nm, similar to the usual TFT, which will not contribute much.
In this way, the semiconductor region, i.e., source/drain can be available though it is thin. This is attributed to that the said region is high in an impurity concentration, and also is in a good crystalline state. In other words, the features of the present invention have been brought about by the laser annealing or the laser doping. Also, the masks used in the present invention are not needed to remain, after completion of TFT, then the unevenness of TFT can be exceedingly reduced.
In accordance with the present invention, a practical P-channel TFT (hereinafter referred to as PTFT), which could not be prepared by a conventional technology in addition to N-channel TFT (hereinafter referred to as NTFT) prepared mainly with a usual amorphous silicon TFT, has come to be prepared. That is, PTFT has not been practical so far, because not only the hole mobility in an amorphous silicon of a channel region was small compared with the ion mobility, but also P-type silicon with low enough resistance of source/drain could not be obtained. In accordance with the present invention, however, the resistance of P-type silicon has been able to be as low as that of N-type silicon, and then, PTFT having a practical function of the device has been able to be prepared.
Therefore, it has begun to be able to prepare a complementary type MOS circuit (hereinafter referred to as CMOS circuit), using the amorphous silicon TFT or TFT prepared at low temperatures. Up to the present time, the CMOS circuit has been limited to the high temperature prepared TFT formed on a quartz substrate at 1000xc2x0 C. or higher, or the middle temperature prepared TFT formed on a non-alkali glass substrate at around 600xc2x0 C. It has been so far considered that CMOS circuit can not be obtained, using TFT prepared at around 350xc2x0 C. of the maximum process temperature.
The instance is shown in FIG. 3. In the same way as indicated in FIG. 1, the gate electrode of NTFT 302 and the gate electrode of PTFT 303 will be formed on the substrate 301, using the first mask. Then if necessary, the gate electrode surface will be oxidized by an anode oxidation method, and thereafter the gate insulating film 304 will be formed. Further, the semiconductor regions of NTFT 305 and PTFT 306 will be formed using the second mask.
The better crystallized state of a semiconductor is, the bigger mobility of PTFT will be obtained. It is not preferable that the mobility between NTFT and PTFT is so different from each other, in order to function as CMOS. Though PTFT with a big mobility can be gotten by enabling a film forming temperature to be higher, it can not be excessively raised under such condition as substrate restrictions. But in case where a film is formed at around 350xc2x0 C. of a substrate temperature using such polysilane as disilane or trisilane, PTFT, which is seemingly amorphous but its mobility is about one over several of that of NTFT, will be gotten. Also, it is possible that an annealing will be carried out in an atmosphere of hydrogen at 300 to 350xc2x0 C. for 24 hr. or more, after the film forming by CVD plasma method.
Then, the silicon nitride films 307 and 308 will be patterned using the third mask. Of course, as mentioned before, this silicon nitride mask may be formed in a self-alignment, by the back exposure using a gate electrode as a mask. In this case the third mask is not necessary. The cross-sectional view of thus obtained device is shown in FIG. 3(A).
After that, the photoresist mask 309 will be formed in the region of PTFT using the fourth mask, and laser beams will be radiated in an atmosphere of phosphine PH3 as illustrated in FIG. 3(B). Thereby the impurity region 310 of NTFT (left side) will be formed. Further, the photoresist mask 311 will be formed in NTFT region using the fifth mask, and laser beams will be radiated in an atmosphere of diborane B2H6 as illustrated in FIG. 3(C), to form an impurity region 312 of PTFT (right side). In each laser doping process, laser beams will be absorbed into the silicon nitride mask, and then the channel forming regions 313 and 314 will not be crystallized.
Then, as shown in FIG. 3(C), the metal wirings (aluminum etc.) 315, 316, 317 will be formed by a known metal wiring technology, thus being formed CMOS circuit consisting of NTFT 318 and PTFT 319. In the above processes, six sheets of mask are used, but one sheet of mask will be reduced, if the back exposure technology is used in case of the preparation of silicon nitride masks 307, 308. The doping process will be also conducted by a known ion implantation or ion doping method. In case where an impurity region is formed by an ion implantation or an ion doping method, which is capable of a delicate controlling of an impurity concentration, it is possible to firstly form an impurity region of either conductivity type in all of TFT, and then to form an inverse conductivity type only in a specific TFT, not separately forming both impurity regions of NTFT and PTFT. In this case, one sheet of mask will be further reduced. This method, however, can not be applicable to the laser doping, on account of its difficulty in an impurity controlling.
If such method is desired to be done by the laser doping, it will be performed as follows; Firstly a certain conductivity type impurity region will be formed toward all of TFT, setting a substrate temperature at a little low, and then an inverse conductivity type doping will be effected only to the specific TFT, raising the substrate temperature. This comes from the reason that the higher the substrate temperature is, the more concentration of an impurity will be doped.
In the present invention, especially regarding the laser doping, such method as is shown in FIG. 4 can be also used. This method conducts the doping in a self-alignment, using the gate electrode as a mask and radiating laser beams from the back. First of all, in the same way as in the case of FIG. 1, the gate electrode 402 will be formed using the mask 1 on the substrate 401 which passes laser beams. If necessary, its oxide 403 will be formed, and also the gate insulating film 404 will be formed. Then, the semiconductor region 405 will be patterned using the mask 2. [FIG. 4(A), (B)].
Next, laser beams will be radiated from the back of substrate. This time, the laser beams pass in parallel in the substrate as illustrated in FIG. 4(C), but are refracted in the gate electrode where is uneven, and diffracted in the gate electrode etc. thus resulting in that the parallel passing is injured. Additionally, in such uneven parts, laser beams are more absorbed in the parts (the oxide layer 403 or the gate insulating film 404), where the laser beams pass through, compared with the other parts. As a result, the intensity of laser beams will be extremely lowered on the protrusion part over the gate electrode part, not only by being masked with a gate electrode but also by the above stated complicated phenomena. Then, the laser doping will no longer be carried out, and the initial situation will be kept to become the channel forming region 406. On the other hand, in the other parts, the laser doping will be performed and the impurity region 407 will be formed. Then, it will do well that the metal wiring 409 and the pixel electrode 410 etc. is formed using the masks 3 and 4.
This method is very simple in the process compared with the other methods. Namely, the number of masks is four sheets, similar to the case where the self-aligning process of the back exposure was adopted in the method of FIG. 1. Also the exposure processes for forming a mask (e.g., 106 of FIG. 1) will be reduced by one, different from the method of FIG. 1. As a matter of course, the forming process of silicon nitride film etc. used for a mask is not necessary. Also, the overlapping of source/drain and gate electrode is rare and the parasitic capacity can be suppressed. This is the most remarkable features of this method.
This method, however, is in need of using a transparent substrate against the laser beams. The Corning 7059 glass substrate is an ideal non-alkali glass, but this is not good in the transparency of ultraviolet beams, so that it is not appropriate to conduct a laser doping using an excimer laser. If the Corning 7059 glass is to be used by all means, it is required to use a long wave of laser beams (e.g., argon ion laser or Nd: YAG laser etc.). Moreover, it is also possible to lengthen twice or more as long as the wave, by using a non-linear type optics effect of the excimer laser beams.